In the

    United States Court of Appeals
                 For the Seventh Circuit
                     ____________________
No. 15-2481
JENNIFER LYNN KRIEGER,
                                                Petitioner-Appellant,

                                 v.

UNITED STATES OF AMERICA,
                                               Respondent-Appellee.
                     ____________________

         Appeal from the United States District Court for the
                     Southern District of Illinois.
           No. 3:14-cv-00749-JPG — J. Phil Gilbert, Judge.
                     ____________________

    ARGUED MAY 24, 2016 — DECIDED NOVEMBER 22, 2016
                ____________________

   Before ROVNER, SYKES, and HAMILTON, Circuit Judges.
    ROVNER, Circuit Judge. The first time this appeal was before
us, on direct review, Jennifer Krieger was seeking to vacate
the twenty-year sentence she received when her friend died
after chewing a fentanyl pain patch provided by Krieger. At
that time, she objected to the manner in which the govern-
ment proved that “death resulted” from the distribution of
the drugs—as a sentencing factor by a preponderance of the
2                                                    No. 15-2481

evidence, rather than as an element proved beyond a reason-
able doubt—and also argued that the evidence was insuffi-
cient to support a finding that the victim’s death had occurred
because of the fentanyl. Given the statutory sentencing struc-
ture in place at the time, and the fact that the district court
found, by a preponderance of the evidence, that death had re-
sulted from the distribution, there was only one sentence that
the district court could give, and that was twenty years. The
district court expressed discomfort with its lack of discretion
and the fact that it appeared that Krieger was being sentenced
for homicide despite having been convicted only of distrib-
uting fentanyl—concerns that this court echoed on appeal.
Nevertheless, based on then current law, we found no error
and affirmed the decision of the district court. After Krieger’s
sentencing and after her direct appeal, the Supreme Court is-
sued two decisions that touch on the very issues raised at
Krieger’s sentencing. Consequently, on June 30, 2014, Krieger
filed a petition under 28 U.S.C. § 2255 asking the court to va-
cate, set aside, and correct her sentence based on new Su-
preme Court rules that she argues should be applied retroac-
tively on collateral review.
    The facts below are both abridged and supplemented from
the decision we issued on December 7, 2010, during Krieger’s
first (direct) appeal as reported in United States v. Krieger, 

, 869 (7th Cir. 2010).
    The afternoon before Thanksgiving, 2005, Jennifer Curry’s
mother found her nineteen-year-old daughter dead on a sofa
at the home of Curry’s father. At the scene, investigators
found, among other things, a chewed 100 microgram Dura-
gesic patch. Duragesic is a brand name for a fentanyl skin
patch, a powerful opioid that is delivered across the skin in
No. 15-2481                                                  3

small steady doses over the course of several days to control
pain. It is not meant to be ingested orally nor injected under
the skin, but sometimes is by those who are abusing the drug.
Of course, fentanyl is available only by prescription and, not
surprisingly, Jennifer Curry did not have one. Her friend, Jen-
nifer Krieger, however, had such a prescription and despite
her pain from severe spinal cord and disk problems, she be-
gan selling the patches to others for $50 apiece or, as hap-
pened here, giving them to her friends. On November 22,
2005, Krieger filled her prescription for the patches and later
that afternoon gave one to Curry. Krieger left Curry at around
midnight and another witness saw Curry leave a bar with two
men in the early hours of November 23. Curry arrived at her
father’s home at approximately two o’clock in the morning.
Her mother found her unresponsive at approximately four
o’clock the next afternoon and arriving paramedics deter-
mined that Curry had been dead for some time. At the scene,
the investigators found a hypodermic needle, a small pipe
with burnt residue on it, and two red capsules. Neither the
two red capsules nor the pipe were taken into evidence and
tested. The syringe was not tested until three years later, at
the request of the U.S. Attorney’s office. A medical examiner
found traces of many drugs in Curry’s system, including co-
caine, benzodiazepines, cannabinoids, and Oxycodone, but
concluded that Curry died from fentanyl toxicity.
    A federal grand jury returned a two-count indictment on
January 5, 2006, charging Krieger with distribution of divers
amounts of fentanyl with death resulting, under 21 U.S.C.
§ 841(a)(1) and § 841(b)(1)(c). Krieger conceded that she gave
Curry a patch. She denied, however, that the government
proved sufficiently that Curry’s death resulted from her
abuse of the fentanyl patch.
4                                                    No. 15-2481

    The government, it seems, also quickly realized that its
case for “death resulting” faced some heavy obstacles. In a
strange twist of events, the government’s main witness, the
medical examiner, Dr. John Heidingsfelder, fled the country
under a cloud of suspicion. It seems that Heidingsfelder had
legal problems of his own, including tax and ethics trouble,
and had left the country and set up a practice in the Cayman
Islands. Investigators for the United States Attorney’s office
had been unable to track him down. Heidingsfelder also had
been disciplined by the Indiana Medical Licensing Board for
engaging in a prohibited personal relationship with a patient,
for prescribing medication to his girlfriend/patient, and fail-
ing to keep abreast of current professional theory and prac-
tice. Apparently, Heidingsfelder had engaged in sexual con-
tact with a patient under his care and provided her hydroco-
done and other narcotic drugs. The woman committed sui-
cide after Heidingsfelder terminated the relationship.
    With the main witness unavailable, the government in-
formed the court that it was engaged in good faith plea nego-
tiations. When those negotiations failed, the government re-
turned a one-count superseding indictment in which the
“with death resulting” language of the indictment had been
eliminated. In this superseding indictment, Krieger was
charged only with distribution of divers amounts of fentanyl
in violation of 21 U.S.C. § 841(a)(1) and § 841(b)(1)(C). The fol-
lowing day, Krieger filed a motion to dismiss the indictment
arguing that the unavailability of the doctor who performed
the autopsy presented an incurable confrontation clause and
chain-of-custody problem. The district court denied the mo-
tion but left open the possibility that it would revisit the issue
at a later time. On October 18, 2008, Krieger pleaded guilty to
No. 15-2481                                                   5

the superseding indictment with the specific exclusion that
she was not pleading guilty to causing the death of Curry.
    Although the government dropped the “with death result-
ing” charge from the indictment, which would require the
government to prove those facts beyond a reasonable doubt,
the government nevertheless sought to have Krieger sen-
tenced under a statute that enhances the sentence if the gov-
ernment can prove, by a preponderance of evidence, that
death resulted from the drug’s use. 21 U.S.C. § 841(b)(1)(C).
That statute instructs, “[I]f death or serious bodily injury re-
sults from the use of such substance [such person] shall be
sentenced to a term of imprisonment of not less than twenty
years or more than life.” 

sentence in this
way leads to a significant change in the sentence. Krieger’s
pre-sentencing report set forth a recommended sentencing
range of ten to sixteen months. If the government could prove
by the much more lax preponderance of the evidence stand-
ard that Curry’s death resulted from Krieger’s distribution of
fentanyl, those facts would trigger a mandatory minimum
sentence of twenty years under 21 U.S.C. § 841(b)(1)(C).
    The court held a sentencing hearing on November 18 and
19, 2008, to determine whether the fentanyl had resulted in
the death of the victim. During that hearing, the government
called numerous witnesses, including the previously unavail-
able but subsequently found Heidingsfelder. Krieger called
Dr. Long, a forensic toxicologist who testified regarding prob-
lems with evidence collection and who challenged the deter-
mination of the cause of death.
   The “death resulting” evidence was muddled and slim.
Krieger presented evidence that the investigators and doctor
performing the autopsy focused exclusively or primarily on
6                                                 No. 15-2481

the fentanyl evidence while ignoring evidence related to the
many other drugs in Curry’s system, in particular evidence of
cocaine use. And of course, the misdeeds of the tax cheat,
scofflaw medical examiner hung heavily in the air of the hear-
ing. Nevertheless, after evaluating Dr. Heidingsfelder’s de-
meanor on the stand and his evidentiary presentation, the dis-
trict court concluded that although his testimony about his
personal life was not credible, his testimony as to how he con-
ducted the autopsy and how he arrived at his conclusion as to
the cause of Curry’s death was indeed credible. On appeal we
accepted this factual finding regarding demeanor and credi-
bility, noting that it could not be overturned unless we found
that Heidingsfelder was incredible as a matter of law, which
we could not. 

.
    Heidingsfelder testified that he collected blood and vitre-
ous fluid samples from Curry and sent them to a private and
reputable laboratory in Indianapolis. That lab reported fenta-
nyl in Curry’s blood in the toxic to lethal range. Heidings-
felder found no external traumatic injuries and noted physical
findings consistent with a drug overdose. Although
Heidingsfelder noted needle marks on Curry’s left elbow, and
the lab report indicated the presence of several other drugs in
Curry’s system, Heidingsfelder testified that in his opinion,
based on the facts known about her death, his examination of
her body, and the lab report, Curry’s death was caused by
fentanyl toxicity, and not by the other drugs found in her sys-
tem either taken alone or in combination. The government,
likely because it recognized the problems with its main wit-
ness, Dr. Heidingsfelder, called three other experts: Dr. Mark
LeVaugh, a physician specializing in forensic pathology, Dr.
Michael Evans, a forensic toxicologist, and Dr. Cynthia Mor-
ris–Kukoski, a forensic toxicologist employed by the FBI lab.
No. 15-2481                                                    7

All bolstered Heidingsfelder’s finding of a toxic-to-lethal level
of fentanyl. Krieger, in turn, called Dr. Long, a forensic toxi-
cologist. Long criticized Heidingsfelder for failing to test the
needle marks on Curry’s body, for his choice of location to
draw blood, and for recording an incorrect time of death,
among other things. He also questioned whether blood sam-
ples had been mishandled or placed in vials with the incorrect
preservative. He maintained that the reports from the lab
used by Heidingsfelder and the FBI laboratory reports were
either incomplete or inconsistent with each other. Finally, he
testified that the blood was not tested immediately and there-
fore would not accurately indicate the use of cocaine. As he
testified, “you can have somebody who overdoses on cocaine,
and if the blood sample isn’t taken relatively close and pre-
served properly, all the cocaine breaks down to benzoylec-
gonine and becomes an un-interpretable result.” Petitioner’s
Sep. App. at 33 (R. 165, p.19, Page ID 715). He questioned
whether the failure to investigate and follow up on other pos-
sible causes of death resulted in an incorrect determination of
the cause of death. Nevertheless, he admitted that the lab re-
ports showed the presence of fentanyl in Curry’s system in
amounts that were four times the therapeutic range and thus
potentially lethal. According to Dr. Evans, a toxicologist who
testified, the therapeutic range for fentanyl when used in such
a patch is 1-3 nanograms per mil. Jennifer Curry’s was more
than 13 nanograms per mil. (R. 164, p. 178, page ID 636).
Deaths from fentanyl occur at amounts from 2.2 to 100 nano-
grams per mil. 

chief in charge of the investigation admitted to
inadequate police work in some areas. Officers failed to col-
lect into evidence and send for testing two red capsules in
Curry’s bedroom. The syringe was not tested until three years
8                                                 No. 15-2481

after Curry’s death, but then was found to contain trace
amounts of cocaine. The only items sent to the lab by the po-
lice department for testing were those pertaining to fentanyl.
    On January 16, 2009, the district court issued its order,
finding, by a preponderance of the evidence, that the fentanyl
supplied by Krieger resulted in the death of Curry. In view of
the conflicting evidence as to the cause of Curry’s death, the
court concluded emphatically that the government would not
have been able to prove, beyond a reasonable doubt, that
Krieger’s distribution of fentanyl was the cause of Curry’s
death, had Krieger been charged with that offense. The court
was persuaded, however, that a preponderance of the evi-
dence established fentanyl as the cause of Curry’s death, and
concluded that “the Government has established that it is
more probable than not that Ms. Krieger’s distribution of fen-
tanyl to Ms. Curry resulted in Ms. Curry’s death.” United
States v. Krieger, No. 06-CR-40001-JPG, 

, at *4
(S.D. Ill. Jan. 16, 2009), aff'd, 

(7th Cir. 2010).
    Once the court made the finding, by a preponderance of
the evidence, that death resulted from the fentanyl, it con-
cluded that it was obligated to impose the mandatory statu-
tory minimum under § 841(b)(1)(C) “if death results”—
twenty years. Under Apprendi v. New Jersey, 

(2000), the court could not impose a sentence greater than the
twenty-year statutory limit for distribution of the drug—the
crime to which Krieger pleaded guilty. But, given the twenty
year minimum triggered by the court’s finding, by a prepon-
derance of the evidence, that Curry’s death resulted from the
distribution of fentanyl, neither could the court impose a sen-
tence of less than twenty years. Without death resulting, the
pre-sentence report reflected an advisory Guideline range of
No. 15-2481                                                    9

10-16 months. And, in fact, our previous research indicated
that the average length of incarceration for fiscal years 2006-
2009 under 21 U.S.C. § 841 for distribution of fentanyl without
death resulting was seven months. 

. In
short, in this case the mandatory minimum triggered by facts
found by the court based on a preponderance of the evidence
in sentencing converged with the statutory maximum im-
posed for facts to which Krieger pleaded guilty as an element
of the crime. The result was that the court could impose ex-
actly one sentence—twenty years. The judge had no discre-
tion whatsoever in the choice of sentence.
    The district court at sentencing was uncomfortable, it
seems, with the fact that “Krieger, while convicted of distri-
bution of divers amounts of narcotics, is being sentenced for
homicide.” Krieger, 

, at *6. The court went on
at some length criticizing the sentencing scheme, and de-
clared that “had the Court the discretion to insist that the Gov-
ernment prove beyond a reasonable doubt that the distribu-
tion to which Krieger pleaded guilty resulted in Curry’s death
before imposing the statutory minimum sentence, it would
have exercised that discretion in this case.” 

court made it clear that it was sentencing Krieger to twenty
years as it felt that it had no choice, but specifically noted
       had the Court the discretion to sentence Krieger
       within the entire statutory range for the crime to
       which she pled guilty, it would have applied the
       adjusted base offense level called for in U.S.S.G.
       § 2D1.1(a)(2), finding that the Government has
       established the “death resulting” factor by a
       preponderance of the evidence. As a result,
       Krieger's base offense level would have been 38.
10                                                    No. 15-2481

        The Court would have given her three points off
        for acceptance of responsibility. As a result the
        Court would have found a net offense level of
        35, criminal history category I, with an advisory
        guideline range of 168 to 210 months.

   At the time of Krieger’s first appeal, only facts that in-
creased the penalty of the crime beyond the mandatory statu-
tory maximum had to be submitted to a jury and proved be-
yond a reasonable doubt. 

. Because
the “death resulting” factor only increased the mandatory
minimum, under existing law at the time, it did not have to
be pleaded in the indictment and proved to a jury beyond a
reasonable doubt, but rather could be found by a judge using
a preponderance of the evidence standard just as the district
court did below. 

, 869; Harris v. United
States, 

, 557 (2002), overruled by Alleyne v. United
States, 

(2013). The district court stated, “the
only reason the Court would have to consider “death result-
ing” to be an element of a separate crime rather than a sen-
tencing factor is if it ran afoul of Apprendi.” Krieger, 

, at * 5. But the court noted, correctly at the time, that it
did not.
    All of that changed, however, in 2013, when the Supreme
Court issued its decision in Alleyne v. United States, 

, 2155 (2013). In Alleyne, the Supreme Court held that facts
that increase a mandatory minimum sentence (just like those
that increase the mandatory maximum sentence) must be sub-
mitted to the jury and proved beyond a reasonable doubt. 

were the law of the land at the time that Krieger was
sentenced, therefore, the fact of “death resulting” would have
No. 15-2481                                                     11

had to be submitted to a jury and proved beyond a reasonable
doubt. We know it likely would not have been because the
district court said in no uncertain terms “had the Court ap-
plied the reasonable doubt standard, the Government would
not have met its burden of proof in establishing that Ms.
Krieger’s distribution of fentanyl to Ms. Curry resulted in the
latter’s death.” Krieger, 

, at *4. In short, were
Krieger to be sentenced today she could not be sentenced for
Curry’s death unless the indictment had charged and a jury
had found, beyond a reasonable doubt, that the fentanyl
caused Curry’s death, a finding the district court did not think
possible. Without the mandatory minimum, therefore,
Krieger’s sentence would have been significantly shorter. In
this case, the district court specifically forecast what his sen-
tence would have been without the mandatory minimum—
somewhere within the advisory guideline range of 168 to 210
months. 

    The Alleyne decision, however, by itself, cannot help
Krieger because this circuit has held that the rule announced
in Alleyne is procedural and therefore cannot be applied ret-
roactively on a collateral review, such as this one. Crayton v.
United States, 

, 624 (7th Cir. 2015), cert. denied, 

(2015). But see 

(Williams, J., concurring)
(a procedural rule that requires that criminal defendants must
be convicted beyond a reasonable doubt is essential to or-
dered liberty and therefore should be applied retroactively).
For that reason, this court vacated the portion of its certificate
of appealability that allowed for claims under Alleyne, noting
that it does not apply retroactively on collateral review. (Ct.
App. R. 7).
12                                                  No. 15-2481

    Alleyne, however, was not the only relevant case the Su-
preme Court has issued since Krieger was sentenced. In 2014,
the Supreme Court held that a defendant cannot be liable un-
der the penalty enhancement provision for distribution of a
drug with “death resulting”—the very same provision under
which Krieger’s sentence was enhanced (28 U.S.C.
841(b)(1)(C))—unless such use of the drug was a but-for cause
of death. Burrage v. United States, 

, 892 (2014).
    A new rule announced by the Supreme Court applies to
all cases still pending on direct review, but for cases such as
this one, on collateral review where a final judgment has been
issued, the rule applies only in certain circumstances. Schriro
v. Summerlin, 

, 351 (2004). One such circumstance
is when the new rule is substantive. Welch v. United States, 

, 1264 (citing 

). Rules that
“narrow the scope of a criminal statute by interpreting its
terms, as well as constitutional determinations that place par-
ticular conduct or persons covered by the statute beyond the
State’s power to punish” are substantive and thus apply ret-
roactively. 

–52 (internal citations omit-
ted). “Such rules apply retroactively because they necessarily
carry a significant risk that a defendant stands convicted of an
act that the law does not make criminal or faces a punishment
that the law cannot impose upon him.” 

(internal
quotations omitted). In other words, “a rule is substantive ra-
ther than procedural if it alters the range of conduct or the
class of persons that the law punishes.” 

    Krieger argues that Burrage narrowed the scope of a crim-
inal statute at issue, § 841(b)(1)(C) and placed certain conduct
“beyond the power of criminal law-making authority to pre-
scribe” and thus is a substantive rule. The government agrees,
No. 15-2481                                                                 13

conceding that Burrage is substantive because it defines an es-
sential element of a federal crime in a way that creates the risk
that individuals convicted before the Burrage decision were
improperly convicted of that offense—or in this case, improp-
erly subjected to the sentence mandated for that offense. Just
recently we accepted this same concession by the govern-
ment, that is that Burrage narrowed the scope of the “death
results” enhancement and is thus substantive and applies ret-
roactively. Gaylord v. United States, 

, 505 (7th Cir.
2016). The government has made this same concession before
several other courts which have likewise accepted the conces-
sion. See Ragland v. United States, 

, 1214 (8th Cir.
2015) (“the government conceded … Burrage applies retroac-
tively”); see also, Weldon v. United States, No. 14-0691-DRH,

, at *3 (S.D. Ill. Apr. 17, 2015) (the govern-
ment conceded that Burrage is substantive in nature and is ret-
roactive, but argued that it was inconsequential for other rea-
sons), vacated on other grounds, No. 15-1994, 

(7th Cir. Aug. 24, 2016); United States v. Schneider, 

1197, 1207 (D. Kan. 2015) (“The government concedes, and
the court agrees, that Burrage announces a new substantive
rule of law applicable to cases on initial collateral review.”),
aff’d, No. 15-3247, 

(10th Cir. Nov. 4, 2016);
United States v. Snider, No. 3:07-CR-124-SI, 

,
at *10 (D. Or. Apr. 13, 2016) (accepting government’s conces-
sion that Burrage standard applies).1



1 It is also true that a number of district courts have issued opinions hold-
ing that Burrage does not apply retroactively, although none with any sig-
nificant analysis. Stewart v. United States, 

, 996 (E.D. Wis.
2015) (finding Burrage not retroactive merely by citing other cases finding
same); Alvarez v. Hastings, No. CV214-070, 

, at *1 (S.D.
14                                                           No. 15-2481

    The Supreme Court has not yet declared whether Burrage
should be applied retroactively, but “[d]istrict and appellate
courts, no less than the Supreme Court, may issue opinions”
on initial petitions for collateral review holding in the first in-
stance that a new rule is retroactive in the absence of a specific
finding to that effect by the Supreme Court. Ashley v. United
States, 

, 673 (7th Cir. 2001); see also Butterworth v.
United States, 

, 465 (1st Cir.), cert. denied, 

(2015) (noting appellate court’s authority to decide if Al-
leyne should be applied retroactively); United States v. Swinton,

, 486 (3d Cir. 2003) (“courts of appeals and district
courts may determine whether a novel decision of the Su-
preme Court applies retroactively.” Our reasoning in Ashley
was as follows:


Ga. Oct. 21, 2014) (adopting magistrate judge’s report and recommenda-
tion found at 

at *1, explaining that petitioner was not
allowed to file second or successive § 2255 motion because the Supreme
Court “did not expressly hold that Burrage is retroactive on collateral re-
view.”), appeal dismissed (July 13, 2015); United States v. Bourlier, No.
3:14cv609, 

, at *2 (N.D. Fla. December 1, 2014) (stating
that the Supreme Court has not expressly held that its holding in Burrage
is retroactively applicable to cases on collateral review); De La Cruz v.
Quintana, No. 14–28–KKC, 

, at *6 (E.D. Ky. May 1, 2014)
(“The Court in Burrage did not indicate whether Burrage applies retroac-
tively, and to date, the Court is unaware of any authority that would make
it retroactive” but deeming it unnecessary to decide); Taylor v. Cross, No.
14–CV–304-DRH, 

, at *3 (S.D.Ill. Mar. 26, 2014); Powell v.
United States, No. 3:09–CV–2141, 

, at *2 (D. Conn. Oct. 10,
2014) (finding that Burrage does not apply retroactively because it is only
about the procedural issues of who decides a given question and under
what standard); United States v. Grady, No. 5:10CR0002, 

,
at *4, (W.D. Va. Aug. 12, 2015) (finding Burrage is not retroactive “because
the Supreme Court has not declared it to be retroactively applicable, and
no other court has such authority.”)
No. 15-2481                                                      15

       An initial petition may be filed within a year of
       a decision that is “made retroactively applicable
       to cases on collateral review[.]” A second peti-
       tion, by contrast, depends on “a new rule of con-
       stitutional law, made retroactive to cases on col-
       lateral review by the Supreme Court” (emphasis
       added). Both statutes make it clear that only the
       Supreme Court may issue the new decision. But
       who decides whether that new decision applies
       retroactively? The first formulation (“made ret-
       roactive”) leaves that question open. The sec-
       ond formulation (“made retroactive … by the
       Supreme Court”) answers it. To treat the first
       formulation as identical to the second is not
       faithful to the difference in language. By omit-
       ting the restriction contained in ¶ 8(2), ¶ 6(3) im-
       plies that courts of appeals and district courts
       may “make” the retroactivity decision. Tyler
       concludes that the word “made” in ¶ 8(2) means
       “held.” 533 U.S. at 

. Dis-
       trict and appellate courts, no less than the Su-
       preme Court, may issue opinions “holding”
       that a decision applies retroactively to cases on
       collateral review. The jurisdictional (and prece-
       dential) scope of that holding differs, but it is a
       holding nonetheless.

(emphasis in original).
     Of course we are not bound to accept the government’s
concession when the point at issue is a question of law. Cos-
tello v. BeavEx, Inc., 

, 1061 n.4 (7th Cir. 2016). Nev-
ertheless, that concession seems apt. In Burrage, the defendant
16                                                    No. 15-2481

was found guilty of distribution of heroin with death result-
ing from the use of that heroin. 

. The
victim in that case had been on a drug bender and had multi-
ple drugs from various sources in his body at the time of
death. Burrage’s lawyer argued that the jury had to find that
the victim would not have died but for the use of the heroin
distributed by Burrage. The court ultimately accepted the
government’s jury instruction which required only that the
jury find that the heroin was a contributing cause. The jury
convicted, and the appellate court affirmed his conviction 

Court, however, reversed, holding that, “at least
where use of the drug distributed by the defendant is not an
independently sufficient cause of the victim’s death or serious
bodily injury, a defendant cannot be liable under the penalty
enhancement provision of 21 U.S.C. § 841(b)(1)(C) unless such
use is a but-for cause of the death or injury.” Burrage, 134 S.
Ct. at 892. This holding was critical in Burrage, as no expert
was willing to say that the victim would have died from the
heroin use alone, but it was clear that the heroin certainly con-
tributed to an overall effect that caused the victim to stop
breathing and die. 

890.
    The lower courts that have concluded that Burrage cannot
be applied retroactively have largely relied on the fact that the
Supreme Court has not so held. See footnote 

. But as
we explained extensively above, “[d]istrict and appellate
courts, no less than the Supreme Court, may issue opinions”
on initial petitions for collateral review holding in the first in-
stance that a new rule is retroactive in the absence of a specific
finding to that effect by the Supreme Court. 

. And to the extent that some of those courts thought
that Burrage was just an extension of the procedural rule in
Alleyne, they were simply incorrect. For example, one district
No. 15-2481                                                   17

court concluded that Burrage did not apply retroactively as it
was “only about procedure—i.e., who decides a given ques-
tion, judge versus jury, and under what standard, preponder-
ance versus reasonable doubt.” Powell v. United States, No.
3:09–CV–2141, 

, at *2 (D. Conn. Oct. 10,
2014).
    But this misunderstands the holding of Burrage. Burrage
did reference Alleyne to establish that “[b]ecause the ‘death re-
sults’ enhancement increased the minimum and maximum
sentences to which Burrage was exposed, it is an element that
must be submitted to the jury and found beyond a reasonable
doubt.” 

. This was a foundational is-
sue in Burrage, however, and not its holding. The Burrage
holding is not about who decides a given question (judge or
jury) or what the burden of proof is (preponderance versus
proof beyond a reasonable doubt). It is rather about what must
be proved. The holding of Burrage could not be more clear, for
the Supreme Court set forth its holding, by stating, “We hold
that, at least where use of the drug distributed by the defend-
ant is not an independently sufficient cause of the victim’s
death or serious bodily injury, a defendant cannot be liable
under the penalty enhancement provision of 21 U.S.C. §
841(b)(1)(C) unless such use is a but-for cause of the death or
injury.” 

In other words, the rule announced in Bur-
rage altered the range of conduct that the law punishes.

.
   A hypothetical will help explain why this is so. Suppose a
former law made it a crime to drive while impaired by alco-
hol, a term that was defined to mean that the accused has had
any amount of alcohol in the previous twenty-four hours—
18                                                  No. 15-2481

even one sip. To enforce this law, the police used a breatha-
lyzer that merely indicated the presence or absence of alcohol.
Anyone whose breath registered the presence of alcohol was
guilty of violating the law. Now suppose that the Supreme
Court found that law to be impermissibly overbroad, and in
response, jurisdictions narrowed the scope of their laws to
state that a driver could not be considered to be impaired un-
less her blood alcohol level registered at .08% or greater—a
threshold commonly thought to signal that the driver is im-
paired. This change is a substantive rule change as it alters the
range of conduct that it punishes and, as to defendants con-
victed under the former rule, it carries “a significant risk that
a defendant stands convicted of an act that the law does not
make criminal or faces a punishment that the law cannot im-
pose upon him.” Schriro at 352. A defendant convicted under
the previous law for driving while impaired might have just
a .02% blood alcohol level—an act that the law does not make
criminal.
    Just as in this hypothetical, the Burrage decision narrowed
the scope of the behavior subject to punishment for “death re-
sulting” by requiring that the drug at issue was the but-for
cause of the victim’s death rather than merely a contributing
cause of death. See 

. As Krieger argues
and the government concedes, the case law supports a finding
that Burrage applied a new substantive rule that must be ap-
plied on collateral review.
    Thus even if we set Alleyne aside for purposes of this col-
lateral review, as we must (for it cannot be applied retroac-
tively), in order to find that death resulted for purposes of the
statutory minimum, the fentanyl would have to be shown to
be the but-for cause of Jennifer Curry’s death.
No. 15-2481                                                      19

    Although the government concedes that Burrage applies
retroactively, it nevertheless argues that the district court did
indeed find that the fentanyl was the but-for cause of Curry’s
death. We are not so certain. The application of the Burrage
holding to a pre-Burrage case presents two problems for an
appellate court trying to determine whether the government
proved that Curry would have died “but for” the fentanyl that
Krieger gave to her.
    First, before Burrage, the district court had no reason to
know that “but-for” causation was the requirement. Nor did
the police, the prosecutor, the medical examiner or the de-
fendant herself. In our example above, in a world where any
amount of alcohol in the system is illegal, there would be no
reason for a court to make a factual determination that a de-
fendant had .08% blood alcohol level. A court that has found
that a defendant is “impaired” may have had before it a de-
fendant with .03% blood alcohol level or .15%. And by apply-
ing the new rule in this situation, where the evidence was not
examined to answer the proper question, the defendant faces
a significant risk that she stands convicted of “an act that the
law does not make criminal or faces a punishment that the
law cannot impose upon h[er].” 

.
    More importantly, given the district court’s language, we
cannot say with any certainty that the district court made a
finding that but for the fentanyl that Krieger supplied, Jen-
nifer Curry would not have died. In fact, the district court did
not use the term “but for” anywhere in the order. The order
is awash in confusion about what causation means in this con-
text. This is not due to any fault on the part of the district court
judge, who did a thorough job on a difficult case, but rather it
is due to the lack of clarity about what type of causation the
20                                                 No. 15-2481

government would have to prove to demonstrate that the
death of a particular victim resulted from a drug that had
been supplied by the defendant. Before Burrage, no district
court had any reason to know that it should be focusing on
“but-for” causation when sentencing for “death resulting.” A
perfect example of the murkiness of causation in this context
can be seen in the following paragraph from the district
court’s order at sentencing:
      the statute imposes the minimum sentence of
      twenty years not when the defendant “causes”
      the death at issue, but when the death “re-
      sults” from the distribution. Dr. Christopher
      Long testified at the sentencing hearing that the
      high levels of fentanyl found in Curry's system
      were not necessarily indicative of the cause of
      death. He said:


             For example, we may have a very
             high concentration of a particular
             drug, a fatal level. And then we
             say, “Well, he was hit by a train.”
             Well, of course, you know what
             killed him. It wasn't the drug. It
             was the train.
      However, for purposes of this statutory man-
      datory minimum, the Government need not
      prove that Krieger ‘caused’ Curry’s death, only
      that Curry’s death ‘resulted’ from Krieger's ac-
      tion. Therefore, if the victim in Dr. Long’s hypo-
No. 15-2481                                                  21

       thetical had fallen asleep on the train track be-
       cause of the fentanyl, we would say the train
       “caused” her death, but her death “resulted”
       from the fentanyl.
Krieger, 

, at *6–7 (emphasis ours). In fact, the
district court goes so far as to say that “[t]he statute imposes
on the Government no need to show that the defendant had
any degree of mens rea. It imposes no requirement of foresee-
ability. It imposes no requirement of causation.” 

(emphasis ours).
    If, in fact, the court was applying a standard in which the
government “need not prove that Krieger ‘caused’ Curry’s
death, only that Curry’s death ‘resulted’ from Krieger’s ac-
tion” (as it appears in the quoted language above), then it was
not applying but-for causation as defined in Burrage. Moreo-
ver, the court used this “death resulting” language through-
out its opinion (all emphases ours):
   • “Having weighed all of the evidence presented, the
Court finds that the Government has established that it is
more probable than not that Ms. Krieger’s distribution of fen-
tanyl to Ms. Curry resulted in Ms. Curry’s death.” Krieger, 

, at *4.
   • “However, had the Court applied the reasonable
doubt standard, the Government would not have met its bur-
den of proof in establishing that Ms. Krieger’s distribution of
fentanyl to Ms. Curry resulted in the latter’s death.” 

the Court, upon a finding that the drug dis-
tributed by Krieger resulted in the death of Jennifer Curry, is
mandated by Congress to sentence Krieger to the statutory
minimum of twenty years.” 

22                                                  No. 15-2481

    • “But, had the Court the discretion to insist that the
Government prove beyond a reasonable doubt that the distri-
bution to which Krieger pled guilty resulted in Curry’s death
before imposing the statutory mandatory minimum sentence,
it would have exercised such discretion in this case.” 

    • “Therefore, the Court, having found that the Govern-
ment has proved by a preponderance of the evidence that the
distribution of fentanyl to which Krieger pled guilty resulted
in the death of Jennifer Curry, is mandated by statute under
21 U.S.C. § 841(b)(1)(C) to sentence Krieger to a term of im-
prisonment of twenty years.” 

    On the other hand, the district court also occasionally used
language more closely associated with but-for causation. The
best support for the notion that the district court was looking
at but-for causation comes from the following statement that
the district court made while describing the testimony of Dr.
Heidingsfelder: “[h]e testified that in his opinion, Curry’s
death was caused by fentanyl toxicity. He further testified
that, in his opinion, the other drugs found in Curry’s system
did not cause her death, either taken alone or in combination.”

The district court opinion also includes language
hinting at but-for causation in the following:
       [i]n addition, the Court finds that the evidence
       presented, including the testimonies of Drs.
       LeVaughn, Evans and Morris–Kukoski, and the
       lab reports from AIT and the FBI, support, by a
       preponderance of the evidence, Dr. Heidings-
       felder's conclusion that fentanyl toxicity was the
       cause of Ms. Curry’s death.
No. 15-2481                                                     23

But yet, despite having said this, the district court
summarized its conclusion by saying, “[h]aving weighed all
the evidence presented, the Court finds that the Government
has established that it is more probable than not that Ms.
Krieger’s distribution of fentanyl to Ms. Curry resulted in Ms.
Curry's death.” 

    As we have emphasized, the confusion is not the fault of
the district court, but rather a lack of clarity in the case law at
the time about what type of causation was required. Before
Burrage, it was unclear what level of causation was necessary
and courts were imprecise with the language they used in
finding causation. In fact, after the district court sentenced
Krieger, another panel of this court issued an opinion explain-
ing how muddled the language of causation is in the cases
applying this same “death resulting” statute. United States v.
Hatfield, 

, 947 (7th Cir. 2010) (explaining causation
in 21 U.S.C. § 841(b)(1)(C)). As the court explained in Hatfield:
       Causation is an important issue in many cases
       in a variety of fields of law and has been so for
       centuries. Yet it continues to confuse lawyers, in
       part because of a proliferation of unhelpful ter-
       minology (for which we judges must accept a
       good deal of the blame). In the space of three-
       and-a-half pages in the government's brief, one
       finds the following causal terms: proximate
       cause, actual cause, direct cause, but-for causa-
       tion, contributing causation, contributory cau-
       sation, significant causal connection, sole cause,
       factor in the victims' injuries, concurrent cause,
       meaningful role, possible cause, remote cause,
       and cause in fact. Black's Law Dictionary (8th
24                                                   No. 15-2481

       ed.2004) lists 26 terms in the entry for “cause.”
       The prosecutor was unable at oral argument sat-
       isfactorily to differentiate or explain the causal
       terms listed in his brief, or the three causal terms
       added to the instruction—“a factor that resulted
       in,” “primary cause,” and “played a part.”

In fact, so confusing was the language of causa-
tion, that although the parties in Hatfield agreed that the gov-
ernment had to prove but-for causation, the government’s
lawyer at argument said that the words “played a part” in the
jury instructions referred to “but-for” causation. 

And in Hatfield, the panel set forth several examples of how
confusing the concept of “but-for” causation can be:
       suppose the ingestion of an illegal drug weak-
       ened the victim’s health to the point where he
       later died of another condition that would not
       have killed him had he not ingested the drug.
       Maybe he was healthy until he ingested it, and
       after and because he ingested it his immune sys-
       tem failed and he died from an overdose of
       drugs, obtained from someone else, that would
       not have killed him but for his weakened condi-
       tion. The government’s lawyer said that ingest-
       ing the first drug would not have been a but-for
       cause of the death. But it would have been: had
       the victim not ingested it, he would not have
       died when he did.
***
       Suppose a defendant sells an illegal drug to a
       person who, not wanting to be seen ingesting it,
No. 15-2481                                                     25

       takes it into his bathroom, and while he is there
       the bathroom ceiling collapses and kills him.
       Had he not ingested the drug, he would not
       have been killed. But it would be strange to
       think that the seller of the drug was punishable
       under 21 U.S.C. § 841(b)(1)(C).

panel described “but-for” causation as “an op-
portunistic concept [that] … we attach to a but-for cause (the
better term is “necessary condition,” since most but-for causes
aren’t considered causes at all) that we’re particularly inter-
ested in, often because we want to eliminate it,” that is, it is
defined by the conduct we wish to deter. 

Hat-
field, we noted that just like attempts to define “beyond a rea-
sonable doubt” for a jury, attempts to define the confusing
term “but-for causation” for a jury “often make[] it less rather
than more clear.” 

    The lack of understanding about what level of cause was
required to make a showing of “resulted in death” led to im-
precise comments by courts and lawyers, including this very
court. In sustaining Krieger’s sentence on direct appeal, we
used the term “but for” in a sentence describing the district
court’s findings, despite the fact that the district court had not
used the term “but-for causation” anywhere in its opinion.

(“The district court properly consid-
ered, therefore, whether Curry’s death would not have oc-
curred but for the ingestion of the fentanyl.”) Yet throughout
the remainder of the opinion, the panel on direct appeal spoke
only of the lower court findings that Krieger “caused” Curry’s
death or that death “resulted.” See 

867, 869, 870 (“the
district court issued its order, finding, by a preponderance of
the evidence, that the fentanyl supplied by Krieger resulted in
26                                                  No. 15-2481

the death of Curry.”); (“the court concluded that the govern-
ment would not have been able to prove, beyond a reasonable
doubt, that Krieger’s distribution of fentanyl was the cause of
Curry's death, had Krieger been charged with that offense.
The court was persuaded, however, that a preponderance of
the evidence established fentanyl as the cause of Curry’s death,
and concluded that ‘the Government has established that it is
more probable than not that Ms. Krieger's distribution of fen-
tanyl to Ms. Curry resulted in Ms. Curry’s death.’”); (“Once the
court made the finding, by a preponderance of the evidence,
that death resulted, it concluded that it was obligated to impose
the mandatory statutory minimum under § 841(b)(1)(C) ‘if
death results’—twenty years.”); (“the court found that Cur-
ry's death resulted from the distribution of fentanyl“); (“The
district court was also correct that once it found, by a prepon-
derance of the evidence, that death had resulted when Krieger
distributed the fentanyl, the court was obliged to impose the
mandatory minimum sentence of twenty years.”); (“the dis-
trict court was correct in allowing the government to prove to
the court at sentencing, by a preponderance of the evidence,
that Curry’s death resulted from Krieger’s distribution of fen-
tanyl.”); (“The district court took these failings into account,
but nevertheless found that it was more likely than not that
the fentanyl patch provided by Krieger caused Curry’s
death.”)
     Before Burrage, the district court had no reason to place
“but-for causation” under a magnifying glass and see it as the
linguistic key to a determination of criminal liability for
death. Neither the court nor the parties had the benefit of Hat-
field or Burrage at the time of sentencing and no reason to
know that but-for causation was the minimum level of causa-
No. 15-2481                                                    27

tion necessary to find that Curry’s death resulted from the in-
gestion of the patch distributed by Krieger. Our own isolated
use of the term but-for causation in our decision sustaining
the district court’s sentence on direct appeal can have no sig-
nificance where the district court made no such finding and
the Supreme Court had yet to declare it the requirement. It is
true, of course, that by the time we issued our opinion on di-
rect appeal, our court had decided Hatfield, in which it began
to unravel what is required to prove that “death resulted
from” the use of the controlled substance as stated in 21 U.S.C.
§ 841(b)(1)(C). It was clear after Hatfield that the government
could not include in its jury instructions a statement that the
controlled substance distributed by the defendant has to have
been “a factor that resulted in the death or serious bodily in-
jury,” and that although it “need not be the primary cause of
death or serious bodily injury” it “must at least have played a
part in the death or in the serious bodily injury.” 

. Hatfield clarified that, at a minimum, the govern-
ment had to prove but-for cause. 

But the Hatfield
opinion left unanswered many questions about but-for causa-
tion. In fact, the court was critical of the term “but-for” causa-
tion noting that “but-for cause is not always (in fact not often)
a cause relevant to legal liability,” and, as we pointed out
above, spoke of the confusion surrounding the term and the
difficulties in pinning it down. 

   Putting all of the pieces together: Burrage applies to
Krieger’s case on collateral review, and therefore requires that
the government show that the fentanyl patch, which Krieger
provided to Jennifer Curry was the “but-for” cause of Curry’s
death, that is, ‘that the harm would not have occurred’ in the
28                                                   No. 15-2481

absence of—that is, but for—the defendant’s conduct.” Bur-

–88 (citing University of Tex. Southwestern
Medical Center v. Nassar, 

, 2525 (2013)).
    In its supplemental filing, the government points to evi-
dence from experts who believed that Curry had a lethal level
of fentanyl in her system and therefore concluded that the
cause of Curry’s death was fentanyl toxicity. In a world that
did not require but-for causation, this evidence was sufficient
to satisfy the “death resulting” language of the enhancement
by a preponderance of the evidence, as the district court
found. Krieger, 

, at *4. Krieger, however, ar-
gues that law enforcement, medical examiners and the like,
focused only on the fentanyl to the exclusion of the other
drugs in Curry’s system which, in a landscape requiring but-
for causation, makes a difference. As Krieger notes, “if before
any testing or investigation there is a 50% chance a victim
died from Drug A and a 50% chance the same person died
from Drug B, can causation determination be made if only
Drug A was tested?” (Krieger Supplemental Brief at 9).
    The government argues that these factual findings—that
fentanyl caused Curry’s death—are reviewed for clear error
and that this court, on direct appeal, has already upheld the
district court’s factual findings. But this court, on direct ap-
peal, upheld the district court’s factual findings that fentanyl
caused Curry’s death. It did not uphold any factual finding
that fentanyl was the “but-for” cause of Curry’s death. It
could not have done so, just as the district court, at the time
of sentencing, had no reason to know that it must make a find-
ing of but-for causation, because it had neither the guidance
of Hatfield nor Burrage. And even though we had some idea,
by the time of direct review, that, because of Hatfield, the
No. 15-2481                                                      29

standard in this circuit was turning to “but-for” causation, the
Hatfield opinion merely reinforced the notion that the term
“but for” was both ill-defined and oft misused. 

. It was not until Burrage that the Supreme Court
mandated that no defendant could be liable under the penalty
enhancement provision of 21 U.S.C. § 841(b)(1)(C) unless the
use of the distributed substance was the but-for cause of
death, a term it defined by stating “this requires proof that the
harm would not have occurred in the absence of—that is, but
for—the defendant’s conduct.” 

, 892.
    But even if this court had somehow upheld the factual
findings that the government claims it did, the law of the case
doctrine is “not hard and fast.” Kathrein v. City of Evanston, Ill.,

, 685 (7th Cir. 2014). Instead, “a ruling made in an
earlier phase of a litigation controls the later phases unless a
good reason is shown to depart from it.” Tice v. Am. Airlines,
Inc., 

, 853 (7th Cir. 2004). And one good reason to
depart from it is when, as here, there is “a decision of the Su-
preme Court after the first review that is inconsistent with the
decision on that review.” 

(citing Chi.
& N.W. Transp. Co. v. United States, 

, 930 (7th Cir.
1978)). Burrage is, of course, just that case.
    We therefore vacate Krieger’s sentence and remand the
case to the district court for de novo resentencing. Because
Krieger’s sentence is vacated, the district court will be resen-
tencing Krieger on a clean slate. See Pepper v. United States, 

, 507 (2011); United States v. Atkinson, 

,
1223 (7th Cir.1992) (vacation of a sentence results in a “clean
slate” and allows the district court to start from scratch);
United States v. Barnes, 

, 330 (7th Cir. 1991) (same).
As the district court noted in its initial opinion on sentencing,
30                                                  No. 15-2481

it can consider the death of Jennifer Curry, using a prepon-
derance of the evidence standard, as recommended under
United States Sentencing Guideline § 2D1.1, which advises
that a judge apply a higher base level where the offense of
conviction establishes that death or serious bodily injury re-
sulted from the use of the substance. U.S.S.G. § 2D1.1. Imple-
mentation of the mandatory minimum sentence under 21
U.S.C. § 841(b)(1)(C), would require that a jury find, beyond
a reasonable doubt, that the fentanyl in the patch provided by
Krieger was the but-for cause of Curry’s death, and the time
for such a determination has passed. 

;

.
    The decision of the district court is VACATED and the case
is REMANDED for resentencing.